DeepSeek R1 should probably see it's earlier reasoning

I don't know if this would be the correct thing to do, but I'd assume it would be useful.

Currently, DeepSeek only sees the messages it wrote earlier, but not the thought process. So if you ask it something like “come up with a number but don't tell it” and then ask it to tell the number, it will say a different number from the one it chose in it's earlier thought process.

It would probably be nice to have it remember it's own thoughts? Again though, not sure if this would maybe make the experience worse somehow